CORRECTION

Correction: Single cell analysis of adult mouse skeletalmuscle stem cells in homeostatic and regenerative conditions(doi: 10.1242/dev.174177)Stefania Dell’Orso, Aster H. Juan, Kyung-Dae Ko, Faiza Naz, Jelena Perovanovic, Gustavo Gutierrez-Cruz,Xuesong Feng and Vittorio Sartorelli

There was an error published in Development (2019) 146, dev174177 (doi:10.1242/dev.174177).

On p. 6, the definition of TA was given incorrectly.

Corrected:

Tibialis anterior (TA) and extensor digitorum longus (EDL) muscles of two 3-month-old C56BL/6J mice were injured with the myotoxicagent notexin and MuSCs were FACS-isolated 60 h after injury (Fig. 4A, Fig. S3B).

Original:

Transverse abdominal (TA) and extensor digitorum longus (EDL) muscles of two 3-month-old C56BL/6J mice were injured with themyotoxic agent notexin and MuSCs were FACS-isolated 60 h after injury (Fig. 4A, Fig. S3B).

Both the online full-text and PDF versions have been updated.

The authors apologise for this error and any inconvenience it may have caused.

1

© 2019. Published by The Company of Biologists Ltd | Development (2019) 146, dev181743. doi:10.1242/dev.181743

DEVELO

PM

ENT

TECHNIQUES AND RESOURCES STEM CELLS AND REGENERATION ARTICLE

Single cell analysis of adult mouse skeletal muscle stem cellsin homeostatic and regenerative conditionsStefania Dell’Orso1,*,‡, Aster H. Juan2,*, Kyung-Dae Ko2,*, Faiza Naz1, Jelena Perovanovic2,Gustavo Gutierrez-Cruz1, Xuesong Feng2 and Vittorio Sartorelli2,‡

ABSTRACTDedicated stem cells ensure postnatal growth, repair and homeostasisof skeletal muscle. Following injury, muscle stem cells (MuSCs) exitfrom quiescence and divide to reconstitute the stem cell pool and giverise to muscle progenitors. The transcriptomes of pooled MuSCs haveprovided a rich source of information for describing the geneticprograms of distinct static cell states; however, bulk microarray andRNA sequencing provide only averaged gene expression profiles,blurring the heterogeneity and developmental dynamics ofasynchronous MuSC populations. Instead, the granularity required toidentify distinct cell types, states, and their dynamics can be affordedby single cell analysis. Wewere able to compare the transcriptomes ofthousands ofMuSCs and primarymyoblasts isolated fromhomeostaticor regenerating muscles by single cell RNA sequencing. Usingcomputational approaches, we could reconstruct dynamic trajectoriesand place, in a pseudotemporal manner, the transcriptomes ofindividual MuSC within these trajectories. This approach allowed forthe identification of distinct clusters of MuSCs and primary myoblastswith partially overlapping but distinct transcriptional signatures, as wellas the description of metabolic pathways associated with definedMuSC states.

KEYWORDS: Muscle stem cells, Satellite cells, Single cell RNA-seq,Muscle regeneration, Mouse

INTRODUCTIONGrowth and homeostasis of the skeletal muscle system, a tissue/organ constituting approximately 35% of the human total bodymass (Janssen et al., 2000), rely on tissue-resident skeletal musclestem cells (MuSCs). Following a period of postnatal proliferationleading to a muscle mass increase of several-fold (White et al.,2010), adult MuSCs enter quiescence (Chargé and Rudnicki, 2004).In response to acute or chronic muscle injury, MuSCs are activated,proliferate and differentiate into either newly formed or pre-existingmuscle fibers to ensure compensatory muscle regeneration andrepair (Brack and Rando, 2012; Comai and Tajbakhsh, 2014).Whereas quiescent and activated MuSCs express the transcriptionfactor Pax7 and accumulate its cognate protein (Seale et al., 2000;Oustanina et al., 2004), transcripts of the myogenic regulators Myf5

and MyoD are present in quiescent MuSCs (Beauchamp et al.,2000; Crist et al., 2012; Liu et al., 2013; Ryall et al., 2015b; deMorree et al., 2017) but their corresponding proteins are detectedonly after MuSCs become activated (Zammit et al., 2002). MuSCsretain regenerative capabilities when isolated from donors andtransplanted into recipient hosts (Sherwood et al., 2004; Collinset al., 2005; Montarras et al., 2005; Kuang et al., 2007). In onestudy, a single MuSC transplanted into the muscles of miceextensively proliferated giving rise to tens of thousands of cellscontributing to muscle fibers. In contrast, transplanted primarymyoblasts were ineffective at engrafting and proliferating (Saccoet al., 2008). Failure of MuSCs to efficiently undergo cell divisionand regenerate is observed during aging (Brack and Rando, 2007;Brack et al., 2007; Kuang et al., 2007; Chakkalakal et al., 2012;Sousa-Victor et al., 2014; Bernet et al., 2014; Cosgrove et al., 2014;Price et al., 2014; Tierney et al., 2014; Sacco and Puri, 2015; García-Prat et al., 2016; Lukjanenko et al., 2016). Moreover, dystrophindeficiency compromises the ability of MuSCs to undergoasymmetric division resulting in inefficient generation ofmyogenic precursors (Dumont et al., 2015; Feige et al., 2018).The transitions experienced by MuSCs during development andregeneration are regulated by partially overlapping and dynamictranscriptional programs (Buckingham and Relaix, 2007; Sincenneset al., 2016; Chang et al., 2018). Bulk transcriptome analysesof MuSCs have contributed important mechanistic insights byproviding static snapshots of averaged gene expression(Pallafacchina et al., 2010; Liu et al., 2013; Sousa-Victor et al.,2014; Yennek et al., 2014; Ryall et al., 2015b; Alonso-Martin et al.,2016; Aguilar et al., 2016; Machado et al., 2017; van Velthovenet al., 2017; Pala et al., 2018). The granularity required to identifydistinct cell types, states, and their transcriptional dynamics can,however, be provided by single cell analysis.

Here, we use droplet-based single cell RNA-seq (scRNA-seq) tocapture the transcriptional state of thousands of individual MuSCsand primary myoblasts. Using computational approaches, we inferdevelopmental dynamic trajectories of homeostatic and regenerativeadult MuSCs and of primary myoblasts, and describe the relativetranscriptional changes, including those relative to metabolicpathways, associated with discrete cell state transitions.

RESULTSscRNA-seq permits the identification of cell populationsof the mouse adult hindlimb skeletal musclesWe performed scRNA-seq to identify and enumerate cellpopulations of mouse adult hindlimb skeletal muscles. Muscles of3-month-old C56BL/6J mice were dissociated into single cells byenzymatic digestion and live cells were isolated by fluorescence-activated cell sorting (FACS) (Fig. 1A). Cells obtained from twomice were sequenced separately. The two samples were tested forsimilarity and merged for further analysis (Fig. S1A). After qualityReceived 26 November 2018; Accepted 7 March 2019

1Genome Technology Unit, National Institute of Arthritis and Musculoskeletal andSkin Diseases (NIAMS), NIH, Bethesda, MD 208292, USA. 2Laboratory of MuscleStem Cells and Gene Regulation, National Institute of Arthritis and Musculoskeletaland Skin Diseases (NIAMS), NIH, Bethesda, MD 208292, USA.*These authors contributed equally to this work

‡Authors for correspondence (stefania.dell’orso@nih.gov;vittorio.sartorelli@nih.gov)

S.D., 0000-0002-1306-2820; V.S., 0000-0002-9284-3675

1

© 2019. Published by The Company of Biologists Ltd | Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

control, we retained 4414 cells for downstream analysis. Onaverage, we detected 1221 gene transcripts in each individual cell(156,277 mean reads per cell) (Fig. S1B). These cell and transcriptnumbers are comparable to those recently reported by the TabulaMuris Consortium for droplet scRNA-seq in limb muscle (TabulaMuris Consortium et al., 2018). The sequencing data were analyzedby t-distributed stochastic neighbor embedding (t-SNE) (Satijaet al., 2015). Unsupervised clustering of the quality-controlled cells

of the two combined mouse samples returned nine cell clusters,transcriptomes of which were related to fibroadipogenic precursors(FAPs), tenocyte-like cells, smooth muscle cells, endothelial cells,adaptive and innate immunity cells (B- and T-lymphocytes,macrophages, monocytes), and a small cluster expressing genespresent in mature skeletal muscle (e.g. creatine kinase muscle, Ckm)(Fig. 1B). These findings are similar to those reported in otherstudies (Verma et al., 2018; Giordani et al., 2018 preprint).

Fig. 1. Identification of distinct cell populations in mouse adult skeletal muscle. (A) Scheme of muscle dissection, single cell FACS isolation and scRNA-seq. (B) Graph-based clustering of FACS-isolated single cells identifies distinct clusters corresponding to different cell populations. (C) Heatmap representing thetop 20 most variably expressed genes between cell clusters identified. Colors and numbers correspond to the cell clusters shown in B. (D) Expression ofrepresentative genes in distinct cell clusters. (E) Clustering and t-SNE plot of freshly FACS-isolated single cells with MuSC characteristics. (F) Kernel densityestimation of Pax7 and Myod transcripts in MuSC cluster 1 and cluster 2. (G) Heatmap representing the top 50 most variably expressed genes between the twoMuSC clusters.

2

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

Visualization of the top 20 most variably expressed genes betweencell clusters documented distinct transcriptional programs of the nineclusters (Fig. 1C) and expression of known cell lineage-enriched or-specific transcripts was observed in each of the cell clusters(Fig. 1D). Of the total 4414 cells, 217 (5%) revealed a geneexpression pattern that could be assigned to MuSCs. Within thiscluster, cells were observed to express variable levels of the MuSCmarkers vascular cell adhesion molecule 1 (Vcam1),Myod (Myod1),Myf5 and desmin (Des) (Chargé and Rudnicki, 2004) (Fig. S1C).Skeletal muscle dissociation and FACS isolation affect a subset oftranscripts in a subpopulation of MuSCs (van den Brink et al., 2017).These manipulations, necessary to obtain MuSCs that can beemployed in transplantation protocols (Montarras et al., 2005;Sacco et al., 2008), decrease expression of transcripts representedin niche-associated quiescent MuSCs, such as that of the paired box/homeodomain Pax7 (Seale et al., 2000), and promote accumulationof transcripts enriched in activated MuSCs, such as that encodingMyoD (Machado et al., 2017; van Velthoven et al., 2017). Consistentwith these observations, t-SNE analysis of theMuSC cluster revealedthe presence of two juxtaposed clusters (clusters 1 and 2; Fig. 1E).Using non-parametric kernel density estimation (KDE), wedetermined the probability density function of cells expressingPax7 and Myod in MuSC cluster 1 and cluster 2. In cluster 1, KDEidentified two cell populations: one with low, the other with highPax7. Cluster 2 contained only low-expressing Pax7 cells. KDE forMyoD transcripts identified two cell populations in cluster 2: onewith lower, the other with higher Myod. Cluster 1 was enriched forlow-expressing Myod cells (Fig. 1F). A heatmap illustratingexpression of the top 50 most variable genes in the two MuSCsubpopulations is shown in Fig. 1G and the corresponding data arereported in Table S1. Overall, scRNA-seq of mononucleated cellsobtained from dissociated hindlimb skeletal muscles permitted theidentification of distinct cell lineages, including MuSCs.

scRNA-seq of FACS-purified muscle stem cellsMuSCs derived from hindlimb muscles of two 3-month-old C56BL/6J mice were prospectively FACS-purified as described (Liu et al.,2015) [VCAM1+/CD31 (PECAM1)−/CD45 (PTPRC)−/Sca1(Ly6a)−] and immediately sequenced (Fig. 2A, Table S1). The twosamples were tested for similarity and merged for further analysis(Table S1, MuSCs1 versus MuSCs2 expression sheet; Fig. S2A,B).After quality control, we retained 3081 MuSCs for downstreamscRNA-seq analysis. On average, we detected 994 expressed genes ineach individual MuSC (Fig. S2C). Using the Chromium platform(10x Genomics), 50,000-70,000 mean reads per cell are generallysufficient to approach saturation, and primary cells with lowRNA content and complexity, such as MuSCs, may require lesssequencing to achieve saturation reads of 80-90% (https://kb.10xgenomics.com/hc/en-us/articles/115002474263-How-much-sequencing-saturation-should-I-aim-for-; Zhang et al. 2019).To approximate sequencing saturation, we obtained 236,841 mean

reads per cell (Fig. S2C). t-SNE analysis identified two MuSCclusters (Fig. 2B). Transcripts of the cell cycle inhibitor Cdkn1a werepresent in both clusters whereas those of cyclin E were not detectedsuggesting that all FACS-isolated MuSCs remained G0-arrested(Fig. 2C). Differentially expressed genes in the two MuSC clustersincluded growth arrest-specific 1 (Gas1), the Notch targetHes1,Pax7and the calcitonin receptor (Calcr) in one cluster (MuSCs close-to-quiescence, MuSC cQ) (Fig. 2C,D, Table S1). Genes associated withearly activation, such as Myod, Myf5 and ribosomal genes, wereinstead enriched in the other MuSC cluster (MuSCs early-activation,MuSCeA) (Fig. 2D, Table S1). TheMuSC cQ cluster comprised 975

cells (975/3081; 32% of total MuSCs) and theMuSCeA cluster 2108cells (2108/3081; 68% of total MuSCs). Gene ontology (GO)analysis confirmed that the two MuSC clusters are transcriptionallydistinct (Fig. S2D, Table S1). GO terms related to resistance to stressand response to unfolded protein, cell cycle arrest and circadianrhythm were enriched in the MuSC cQ cluster whereas termsindicating activation of ribosome biogenesis, mRNA processing,translation, and protein stabilization were enriched in the MuSC eAcluster (Fig. 2D, Fig. S2D, Table S1). Transcriptome comparisonbetween FACS-isolated MuSCs (ex vivo MuSCs) and in vivoquiescent MuSCs indicated that the MuSC ex vivo transcriptomeremains largely reflective of the in vivo transcriptome (van Velthovenet al., 2017). However, another study has reported markedtranscriptional differences between MuSCs FACS-isolated fromunfixed or paraformaldehyde (PFA)-fixed muscles (Machado et al.,2017). To evaluate the transcriptional state of MuSC cQ and MuSCeA clusters, we compared their respective transcriptomes with that ofMuSCs isolated from PFA-fixed muscles to remove genes for whichtranscription is affected by muscle dissection and FACS isolation(Machado et al., 2017). This analysis revealed that although MuSCeA andMuSCs derived from PFA-fixed muscles shared only 1.8% oftranscripts, the percentage increased to 23% inMuSC cQ (Fig. 2E,F).

Myod transcripts are present in quiescent MuSCs but theircorresponding protein can be detected only in activated MuSCs. Todetermine whether FACS-isolated MuSCs cells translated MyodmRNA into the corresponding protein, we captured 217 MuSCs forsingle cell western blot analysis. Every MuSC was positive for thehistone H3 protein. However, MyoD could be detected in onlyseven out of 217 MuSCs (3.2%) (Fig. 2G). Thus, although tissueand cell manipulations induce transcriptional modifications, theseare not immediately followed by MuSC activation as indicated bythe paucity of freshly isolated MuSCs expressing detectable levelsof the MyoD protein.

scRNA-seq of freshly isolated MuSCs and primary myoblastsidentifies discrete transcriptional programsWe wished to compare scRNA-seq profiles of freshly isolatedMuSCs, which have not fully committed to the muscle cell lineage(Chargé and Rudnicki, 2004), with those of committed primarymyoblasts (PMs). To this end, PMs isolated from the hindlimbmuscles of 3-month-old C56BL/6J mice were cultured in growthmedium before scRNA-seq. After quality control, we retained 4429PMs for downstream analysis. On average, we detected 4933expressed genes in each individual PM (Fig. S3A). To compareMuSCs and PMs directly, we performed t-SNE analysis andcomputationally merged their corresponding scRNA-seq datasets(Fig. 3A, Fig. S2E,F). t-SNE analysis segregated the transcriptomescorresponding to MuSCs and PMs into five clusters: two belongingto MuSCs and three to PMs (Fig. 3A, Fig. S2G,H). Analysis of thetop 20 genes differentially expressed in each of the five clusters isshown in Fig. 3B. Cells present in both PM clusters (PM1 and PM2)expressed transcripts of cyclins Ccnd1 and Ccnd2 (Fig. 3C). Cells ofthe PM1 cluster were enriched for cell cycle regulators of the G1/Sphase (Table S2). A cluster of PMs containing cells that hadactivated transcription reflective of differentiation (differentiatingmyocytes, DM) was characterized by myogenin expression(Fig. 3C). In addition to being negative for Ccnd1 and Ccnd2, thecluster corresponding to MuSC cQ expressed growth arrest-specific1 (Gas1) and the cell cycle inhibitors Cdkn1a and Cdkn2d, whereasMyf5 was preferentially detected in the cluster corresponding toMuSC eA (Table S2). In addition to terms related to the cell cycle,GO analysis of the PM clusters revealed an enrichment for terms

3

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

related to metabolic pathways (oxidative and glycolytic enzymes,serine biosynthesis enzymes, enzymes of the pentose phosphatepathway), proteostasis (proteome quality and homeostasis), andactivation of translation (Fig. 3D, Table S2).To obtain temporal resolution of MuSCs and PMs, we employed

the pseudotemporal ordering algorithm Monocle2 (Qiu et al.,2017). This algorithm permits single cells to be ordered

informatically along trajectories reflecting their progressiontowards differentiation (Trapnell et al., 2014). Cells located in theinitial phase of the trajectory (start of pseudotime) correspond toMuSC cQ, as indicated by expression of Pax7 and the Notch targetHes1 (Fig. 3E). Without an apparent hiatus separating them from theinitial phase, cells located further on the trajectory initiate Myf5transcription and upregulateMyod (MuSC eA). PMs were identified

Fig. 2. Transcriptional characterization of FACS-isolatedMuSCs. (A) Scheme of MuSC FACS isolation and scRNA-seq. (B) Graph-based clustering of FACS-isolated MuSCs (VCAM1+/CD31−/CD45−/Sca1−) identifies two clusters: MuSCs close-to-quiescence (cQ) and MuSCs early activation (eA). (C) Expressionpattern of the cell cycle inhibitor genes Cdkn1a, cyclin E (Ccne1), growth-arrest 1 (Gas1) and the Notch target Hes1. (D) Heatmap of the top 50 mostvariably expressed genes between MuSC cQ and MuSC eA. (E) Percentage of transcripts shared between MuSC eA, MuSC cQ and PFA-fixed t0 MuSCs(Machado et al., 2017). (F) Gene list of transcripts shared between MuSC eA, MuSC cQ and PFA-fixed t0 MuSCs (Machado et al., 2017). Prkcdbp, Cavin3; Sdpr,Cavin2; Selm, Selenom; Sepp1, Selenop. (G) Single cell western blot analysis of 217 freshly FACS-isolatedMuSCs with histone H3 andMyoD antibodies. MyoD+

cells are encircled in red.

4

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

as a distinct group of cells expressing transcripts involved in cellcycle progression (cyclin E and G1/S checkpoint), chromatinremodeling complexes, and regulators of translational initiations

(eukaryotic initiation factors, eIFs) (Fig. 3E, Table S2). PMsbranched off in two distinct trajectories. One trajectory identifiedcells progressing towards differentiation, as indicated by activation

Fig. 3. Clustering and pseudotemporal trajectories identify transcriptional dynamics of MuSCs and PMs. (A) Graph-based clustering of FACS-isolatedMuSCs and PMs showing MuSCs close-to-quiescence (cQ), MuSCs early activation (eA), primary myoblasts cluster 1 (PM1) and cluster 2 (PM2), anddifferentiating myocytes (DM). (B) Heatmap of the top 20 most variably expressed genes between MuSC cQ, MuSC eA, PM1, PM2 and DM clusters.(C,D) Expression patterns of the cyclin genes Ccnd1, Ccnd2, and myogenin (C) and of glycolytic (Eno1, Pkm, Ldha), mitochondrial (Mdh2, Etfa, Etfb), pentosephosphate (Pgd, Pgls), translation initiation factors (Eif3c, Eif4ebp1), ribosomal (Rpl22) and ubiquitin-ligase (Stub1) genes (D). (E) Pseudotime single celltrajectory reconstructed by Monocle2 for MuSC cQ, MuSC eA and PMs. Pax7 and Hes1 are enriched in MuSC cQ andMyf5 andMyod in MuSC eA. PMs branchoff in differentiatingmyocytes (DM) expressingmyogenin (Myog), troponin (Tnnt2) and the cell cycle inhibitorsCdkn1a/b and in PMs that fail to enter differentiationand maintain expression of the cyclins Ccnd1 and Ccdn2.

5

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

of myogenin and the muscle structural transcript Tnnt2 (troponinT2) (Fig. 3E). Genes encoding cyclins were downregulated and cellcycle inhibitors Cdkn1a/b upregulated in cells located within thisbranch (Fig. 3E, DM). Another trajectory stemming out of the PMnode identified cells that retained Ccnd1 and Ccnd2 withoutdetectable myogenin expression (Fig. 3E). This branch correspondsto proliferating myoblasts that have not entered the differentiationprogram and may contain a self-renewal subpopulation of described‘reserve cells’ (Qiu et al., 2017; Cacchiarelli et al., 2018).

Single cell trajectories of homeostatic, injured MuSCs,and PMsTo evaluate how the MuSC transcriptome is impacted by muscleinjury, we generated and compared scRNA-seq datasets ofhomeostatic (uninjured) and regenerating MuSCs. Tibialis anterior(TA) and extensor digitorum longus (EDL) muscles of two3-month-old C56BL/6J mice were injured with the myotoxicagent notexin and MuSCs were FACS-isolated 60 h after injury(Fig. 4A, Fig. S3B). After quality control, we retained 3550 MuSCsfor downstream scRNA-seq analysis. On average, we detected 1336expressed genes in each individual MuSC (248,914 mean reads per

cell; Fig. S3C). The two samples were tested for similarity andmerged for further analysis (Fig. S3D,E, Table S2 60 h Inj1 versus60 Inj2 sheet). t-SNE analysis of injured MuSCs identified threeclusters, MuSCs 60 h Cl1, Cl2 and Cl3, distinct from those ofhomeostatic MuSCs (Fig. 4A, Fig. S3G,H).

GO analysis revealed activation of common transcriptionalprograms in the three injured MuSC clusters (Fig. S3F, Table S2).Translational regulators were enriched in all injured MuSCs (Eif3c,Eif4ebp1), which had also activated Ccnd1 transcription (Fig. 4B,Fig. S3F, Table S2). Cells located in one of the injuredMuSC clustersexpressed myogenin and were enriched for the terms ‘musclecontraction’ and ‘myotube differentiation’, indicating that they hadentered differentiation (Fig. 4B, Table S2). As noted in bulktranscriptome analyses of injured or culture-activated MuSCs(Pallafacchina et al., 2010; Liu et al., 2013; Ryall et al., 2015b;Pala et al., 2018), expression of genes involved in oxidation-reductionprocesses, glycolysis, tricarboxylic acid (TCA) cycle, mitochondrialgenes, serine biosynthesis and pentose phosphate pathways was alsoenriched in injured MuSCs (Fig. 4C, Fig. S3F, Table S2). Amongthem, we found profilin 1 (Pfn1), deletion of which induces a switchfrom glycolysis to mitochondrial respiration in hematopoietic stem

Fig. 4. Clustering and single cell trajectories ofhomeostatic MuSCs, injured MuSCs, and PMs.(A) Scheme of MuSC FACS isolation 60 h after muscleinjury, scRNA-seq, and graph-based clustering ofFACS-isolated homeostatic and MuSCs 60 h afterinjury. Homeostatic MuSCs (in gray) are composed oftwo clusters, cQ and eA, and MuSCs 60 h of threeclusters, 60 h Cl1, Cl2 and Cl3. (B) Expression patternfor Ccnd1, translation initiation factors (Eif3c,Eif4ebp1), ribosomal genes (Rplp1, Rpl26), andmyogenin (Myog). (C) Expression pattern of glycolytic,TCA and mitochondrial genes. (D) Pseudotime singlecell trajectory reconstructed by Monocle2 forhomeostatic MuSCs, injured MuSCs (60 h Inj) andPMs. Transcripts enriched in distinct cell states areindicated. (E) Pseudotemporal heatmap showing geneexpression dynamics for metabolic genes. Genes (row)are clustered and cells (column) are ordered accordingto pseudotime.

6

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

cells (Zheng et al., 2014), and Park7, encoding a redox-sensitivechaperone protecting neurons against oxidative stress and involved inearly-onset Parkinson disease, deletion of which promotes glycolysisin skeletal muscle (Shi et al., 2015). Expression of genes known toregulate metabolic pathways and RNA metabolism (Rplp1, Rpl26)(Fig. 4B, Fig. S3F) and expression of nucleophosmin 1 (Npm1)(Fig. S3I), which promotes RNA metabolism and is essential forembryonic development (Colombo et al., 2005) (Fig. S3F), werealso increased.Pseudotemporal ordering aligned individual homeostatic, injured

MuSCs, and PMs, along a linear trajectory. Homeostatic MuSCswere located at one end and PMs were found at the opposite endof this trajectory. Injured MuScs were temporally distinct fromhomeostatic MuSCs and extended towards PMs (Fig. 4D, Fig. S4A).Correct cell ordering was confirmed by plotting genes known to beexpressed in the three different cell states (Fig. 4D). Next, wegenerated heatmaps of pseudotemporal expression patterns tovisualize metabolic pathways comprehensively in homeostaticMuScs, injured MuSCs, and PMs (Fig. 4E, Fig. S4B). Expressionof mitochondrial genes progressively increased from homeostaticthrough injured MuSCs to PMs, serving as a molecular correlate ofthe increased mitochondria observed in activated and culturedMuSCs (Tang and Rando, 2014; Ryall et al., 2015b).Concordantly, genes of the TCA cycle and the electron transportchain were increased in injured MuSCs and in PMs. With theexception of Hk2 (Pala et al., 2018), glycolysis and pentosephosphate pathways were also activated in injured MuSCs and inPMs (Fig. 4E, Table S2). As previously observed, expression of asubset of oxidative genes (Fabp4, Pnpla2, Acaca, Lipa, Acsl1) waselevated in homeostatic MuSCs and decreased in injured MuSCs andin PMs (Ryall et al., 2015b) (Fig. 4E).

Dynamics ofmetabolic pathway connectivity in homeostaticMuSCs, injured MuSCs and proliferating myoblastsAnalysis of average gene expression confirmed pseudotemporalexpression patterns indicating that transcripts of metabolic enzymeswere increased in both injured MuSCs and proliferating PMscompared with homeostatic MuSCs (Fig. 5A). To evaluate thetranscriptional dynamics of different metabolic pathways inhomeostatic MuSCs, injured MuSCs and PMs, we generatedheatmaps of genetic interactions based on the use of pairwisematrices with Euclidean distance. In such matrices, gene proximityindicates co-expression whereas gene distance indicates absence ofco-expression. Grouping genes into clusters based on the similaritybetween their expression profiles allows evaluation of the presence offunctionally connected and related modules (Stuart et al., 2003;Nguyen and Lió, 2009). Connectivity mapping indicated that genesof the TCA cycle established an interacting module in homeostaticMuSCs. This module was stabilized, as revealed by increased geneinteractions in injured MuSCs, and extended to encompass themajority of TCA genes in PMs (Fig. 5B). Mitochondrial, oxidativephosphorylation and fatty acid oxidation genes steadily increasedtheir connectivity, both in strength and number, from homeostaticMuSCs to PMs (Fig. 5C,D). In a similar manner, electron transportchain (ETC) genes progressively increased their interaction(Fig. S4C). Glycolytic genes established a well-coordinatednetwork in homeostatic MuSCs (Fig. 5E, left). In injured MuSCs,the connectivity of the glycolytic interaction map transitioned to adifferent conformation characterized by newly acquired intermediategene interactions (∼0.75 on the color key) of Ldha with all theglycolytic genes and strengthened connectivity of Pkm (Fig. 5E,middle). Gene connectivity further increased in PMs (Fig. 5E, right).

Interactions among members of the pentose phosphate pathway alsoprogressively expanded and stabilized in injured MuSCs and PMs(Fig. 5F). Connectivity of genes involved in ribosomal biogenesissteadily increased from MuSCs to PMs and that of protein-foldinggenes underwent dynamic changes aswell. In contrast, the network ofgenes for structural proteins remained relatively constant (Fig. S4C).

This analysis revealed that although expression of metabolicpathway genes is concordantly orchestrated resulting in a progressiveenrichment of all the energy-producing pathways from homeostaticand injured MuSCs to PMs, the dynamics and connectivity of geneswithin the individual metabolic pathways are distinct and identifydiscrete cell states.

DISCUSSIONWe have employed scRNA-seq to identify different cell populationsof adult mouse skeletal muscle with a special emphasis onMuSCs. FACS-isolated MuSCs could be identified with distincttranscriptional transition states. This phenomenon has beenpreviously documented using bulk RNA-seq approaches (Machadoet al., 2017; van Velthoven et al., 2017) and scRNA-seq of 21 and235MuSCs, respectively (van den Brink et al., 2017; Cho and Doles,2017). Here, we extend those findings by analyzing thousands ofMuSCs obtained from homeostatic and notexin-injured muscles andcomparing the scRNA-seq datasets with those obtained fromproliferating committed PMs. t-SNE analysis of FACS-isolatedcells identified two MuSC clusters with overlapping but distincttranscriptomes. Indicating MuSC activation, one of the clusterscontained cells with higher Myod and lower Pax7 transcript levelswhereas the other cluster was composed of cells retaining atranscriptome closer to that of PFA-fixed MuSCs (Machado et al.,2017). Although freshly isolated MuSCs express Myod and Myf5transcripts, their cognate proteins are not detected until cell activation(Zammit et al., 2002). This regulatory step is controlled bymicroRNA 489 (Cheung et al., 2012) and microRNA 31 (Cristet al., 2012), and by the RNA-binding proteins of the tristetraprolins(TTP) family, which destabilize Myod mRNA (Hausburg et al.,2015), and Staufen1, which binds to and prevents translation ofMyodmRNA in quiescent MuSCs (de Morree et al., 2017). Our single cellwestern blot analysis of freshly FACS-isolated MuSCs indicated thatthe MyoD protein could be detected in only a small percentage ofMuSCs (∼3%), suggesting that although Myod transcripts areincreased during FACS isolation of MuSCs, they are notimmediately translated.

By aligning the individual transcriptomes of MuSCs derived fromhomeostatic and injured muscles, and primary committed myoblastsalong a pseudotemporal single trajectory, we could order them, in anunsupervised fashion, and obtain a dynamic description of thetranscriptional events characterizing the progression of the individualcells towards differentiation. As expected, the transcriptomes of close-to-quiescence and early-activation clusters present in freshly isolatedMuSCs partially overlapped but the existing differences werebioinformatically recognized, thus allowing the positioning of thetwo cell clusters in close yet distinct space-time positions along thetrajectory. Sharp transcriptome differences between homeostatic andinjured MuSCs resulted in injured MuSCs occupying apseudotemporal position intermediate between homeostatic MuSCsand PMs. Alternate trajectories identified PMs that successfully enteredterminal differentiation, as indicated by expression of myogenin,muscle structural genes, and cell cycle inhibitors, or that retained atranscriptome indicative of a committed yet undifferentiated state.

Several chaperones were enriched in the MuSC cQ cluster.Carbon dating experiments indicate an average age of 15.9 years for

7

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

cells residing in human skeletal muscle (Spalding et al., 2005).Thus, chaperones may afford a quality control mechanism bymaintaining appropriate protein folding (Hartl et al., 2011). Aconcomitant enriched expression of genes involved in ‘response tounfolded protein’ suggests tight proteostatic regulation in the MuSCcQ cluster (García-Prat et al., 2017). The MuSC eA cluster wasenriched for ribosome biogenesis, RNA splicing, and translationalinitiation factors. These GO terms were also present in regenerativeMuSCs and PMs. Thus, in addition to their post-transcriptionalmodifications (Zismanov et al., 2016), expression of translationalinitiation factors was increased in MuSCs and proliferating PMs.Besides serving as building blocks for muscle growth and repair,MuSCs cross-talk with their immediate cellular environment via

extracellular vescicles containing RNA and proteins, mediatingintercellular communication. During this process, MuSCs secreteexosomes containing miRNAs regulating extracellular matrix(ECM) remodeling (Fry et al., 2017). Coordinated activation of anextensive network of genes controlling extracellular exosomesbiogenesis was observed in MuSC eA, regenerative MuSCs, andPMs supporting a potential role for MuSC-secreted vesicles in ECMremodeling that occurs during muscle repair. Interestingly,transcripts of Ccnd1 were detected in MuSCs cQ (Figs 2 and 3),suggesting a role unrelated to cell cycle regulation for cyclin D1.

Metabolic reprogramming is associated with distinct states ofembryonic and somatic cells (Ryall et al., 2015a). In MuSCs,transcriptional activation of both glycolytic and mitochondrial genes

Fig. 5. Metabolic pathways gene connectivity inhomeostatic MuSCs, injured MuSCs, andproliferating myoblasts. (A) Heatmap of averageexpression for metabolic genes in homeostatic andinjuredMuSCs (60 h Inj), and in PMs. (B-F) Distancematrices indicating gene connectivity (highestconnectivity, red; lowest connectivity, blue) ofmetabolic genes in homeostatic and injured MuSCsand in PMs.

8

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

occurs during cell activation and proliferation (Pallafacchina et al.,2010; Rocheteau et al., 2012; Liu et al., 2013; Rodgers et al., 2014;Ryall et al., 2015b; Pala et al., 2018) and interfering with eitherglycolysis or peroxisomal fatty acid oxidation bears functionalconsequences on cell activation and differentiation (Ryall et al.,2015b; Pala et al., 2018). We took advantage of scRNA-seqgranularity to investigate metabolic reprogramming in MuSCs andPMs. Regenerating MuSCs are enriched for all the main energy-producing pathways, including glycolysis (Liu et al., 2013; Palaet al., 2018). Bioenergetics profiles of MuSCs freshly isolated fromadult mice revealed a positive basal oxygen consumption rate(OCR)/extracellular acidification rate (ECAR) ratio (indicative ofoxidative phosphorylation) (Pala et al., 2018), and cultured MuSCsdisplayed increased ECAR (indicative of glycolysis) comparedwith freshly isolated MuSCs (Ryall et al., 2015b). In addition,comparison of transcripts present in both FACS-isolatedMuSCS andin vivo 4-thiouracil-labeled MuSC transcriptomes revealed anenrichment of transcripts for mitochondrial-associated functions(TCA, electron transport chain, oxidative reduction) (van Velthovenet al., 2017), suggesting oxidative phosphorylation in quiescentMuSCs (Rodgers et al., 2014; Ryall et al., 2015b). By analyzing genenetworks based on pairwise matrices of Euclidean distance, wedetected overall increase of all metabolic pathways investigated fromhomeostatic through injured MuSCs to PMs. In injured muscles,increased glycolysis may generate the glycolytic intermediatesnecessary for the production of new biomass andmetabolites utilizedby histone- and DNA-modifying enzymes whereas oxidativephosphorylation would generate the ATP required for MuSCproliferation (Ryall et al., 2015a; Pala et al., 2018). Indeed,metabolites generated by different metabolic pathways regulate theactivity of chromatin-modifying enzymes known to control differentaspects of MuSC biology (McKinnell et al., 2008; Juan et al., 2011;Kawabe et al., 2012; Liu et al., 2013; Ryall et al., 2015b; Boonsanayet al., 2016; Faralli et al., 2016; Scionti et al., 2017; Das et al., 2017;Tosic et al., 2018; Puri and Mercola, 2012). For instance, theoscillation of intermediary metabolites of the NAD biosyntheticpathways may be relevant for circadian gene expression in MuSCsand skeletal muscle (Nakahata et al., 2009; Ryall et al., 2015b;Solanas et al., 2017; Andrews et al., 2010; Lowe et al., 2018). Lastly,perturbations of specific metabolic pathways induce state changes inMuSCs (Ryall et al., 2015b; Pala et al., 2018), revealing a central roleof metabolism in regulating cell state transitions. scRNA-seq haspermitted the identification of cellular states and a granular definitionof the individual cell transcriptional developmental trajectoriesaccounting for the transition of MuSCs from quiescence toproliferation and differentiation. Additional studies will advanceour understanding of the regulatory mechanisms governing thesetransitions at the individual cell level and provide further insightsinto how metabolic pathways are transcriptionally tuned.

MATERIALS AND METHODSMice and animal careMice were housed in a pathogen-free facility and all experiments wereperformed according to the National Institutes of Health (NIH) Animal Careand Use regulations.

Mononuclear cells and MuSC isolation by FACS from mousetibialis anterior muscleHindlimb muscles from 3-month-old C56BL/6J wild-type mice weredissociated into single cells by enzymatic digestion and live cells wereisolated by FACS. Skeletal MuSCs were sorted following describedmethods (Liu et al., 2015). Briefly, hindlimb muscles from 3-month-oldadult wild-type mice were minced and digested with collagenase for 1 h and

MuSCs were released from muscle fibers by further digesting the muscleslurry with collagenase/dispase for an additional 30 min. After filtering outthe debris, cells were incubated with the following primary antibodies:biotin anti-mouse CD106 (anti-VCAM1, BioLegend 105704; 1:75),PE/Cy7 Streptavidin (BioLegend 405206; 1:75), Pacific Blue anti-mouseLy-6A/E (anti- Sca1, BioLegend 108120; 1:75), APC anti-mouse CD31(BioLegend 102510; 1:75) and APC anti-mouse CD45 (BioLegend103112; 1:75). Satellite cells were sorted by gating VCAM1-positive,Pacific Blue-labeled Sca1-negative, and APC-labeled CD31/CD45-negative cells. SYTOX Green (ThermoFisher Scientific S7020; 1:30,000)was used as a counterstain.

PM culture conditionsPMs derived from the hindlimbs of 3-month-old C56BL/6J wild-type micewere cultured for four passages in HAM’s F10 (Invitrogen) supplementedwith 2.5 ng/µl of bFGF (Invitrogen), 10% horse serum (Invitrogen), andpenicillin and streptomycin (Invitrogen).

Muscle injuryC57/BL6 mice (3-6 months old) were anesthetized using isoflurane(1.5-2.5%) via a vaporizer with anesthetic gas-scavenging equipment. Asmall incision (3-5 mm) was made on the right-side leg to expose the TAmuscle. Avolume of 100 µl notexin (2 µM, Accurate Chemical & Scientific)was evenly delivered into TA and EDL muscles using a 29 gauge insulinneedle. The incision was closed by wound clips. Buprenorphine (100 µl0.1 mg/kg diluted in saline) was delivered subcutaneously at the completionof surgery to alleviate the pain. The mice were put on a separate cleanbedding cage and were monitored until they became fully awake, thenmoved back to their original cages.

Single cell sequencing library construction using 10x GenomicsChromium ControllerCells were collected, washed once with 1×PBS and re-suspended in PBSwith 0.04% bovine serum albumin. Cellular suspensions were loaded on aChromium Instrument (10x Genomics) to generate single cell GEMs. Singlecell RNA-seq libraries were prepared using a Chromium Single Cell 3′Library & Gel Bead Kit v2 (P/N 120737, 10x Genomics). GEM-RT wasperformed in a C1000 Touch Thermal cycler with 96-Deep Well ReactionModule (Bio-Rad; P/N 1851197): 53°C for 45 min, 85°C for 5 min; held at4°C. Following retrotranscription, GEMs were broken and the single-strandcDNAwas purified with DynaBeads MyOne Silane Beads (Thermo FisherScientific; P/N 37002D). cDNA was amplified using the C1000 TouchThermal cycler with 96-Deep Well Reaction Module: 98°C for 3 min;cycled 12 times: 98°C for 15 s, 67°C for 20 s, and 72°C for 1 min; 72°C for1 min; held at 4°C. Amplified cDNA product was purified with theSPRIselect Reagent Kit (0.6× SPRI). Indexed sequencing libraries wereconstructed using the reagents in the Chromium Single Cell 3′ Library &Gel Bead Kit v2, following these steps: (1) end repair and A-tailing; (2)adaptor ligation; (3) post-fragmentation, end repair andA-tailing double sizeselection cleanup with SPRI-select beads; (4) sample index PCR andcleanup. The barcode sequencing libraries were diluted at 3 nM andsequenced on an Illumina HiSeq3000 using the following read length: 26bpfor Read1, 8 bp for I7 Index and 98 bp for Read2.

Data processing and clusteringSequences from the Chromium platform were de-multiplexed and alignedusing CellRanger ver. 2.2.0 from 10x Genomics with default parameters(https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/what-is-cell-ranger). For clustering, filtering, variablegene selection and dimensionality reduction were performed using Seuratver.2.3.4 (Butler et al., 2018) according to the following workflow: (1)Cells with fewer than 200 detected genes or in which a high percentage ofUMIs mapped to mitochondrial genes were excluded from furtheranalysis. (2) Cells with a lower percentage (0.06-0.1%) of UMIs mappedto mitochondrial genes were retained for further analysis. (3) The UMIcounts per million were log-normalized for each cell using the naturallogarithm (Butler et al., 2018). (4) Variable genes were selected using

9

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

thresholds (0.05-0.1) after calculating the binned values from log averageexpression and dispersion for each gene (Butler et al., 2018). (5) Theexpression level of highly variable genes was scaled along each gene, andwe regressed out cell-cell variation in gene expression by the number ofdetected molecules (5000-8000), and mitochondrial gene expression. (6)The scaled data were projected onto a low-dimensional subspace ofprincipal component analysis (PCA) using dimensional reduction. Thenumber of PCA were decided through the assessment of statistical plots(Butler et al., 2018). (7) Cells were clustered using a graph-basedclustering approach optimized by the Louvain algorithm with resolutionparameters and visualized using two-dimensional tSNE. (8) To definethe clusters’ characteristics, we identified markers for all clusters withlow minimum percentage of genes detected and reassigned the namesof the clusters on the basis of the distribution of the specific markers.In addition, using the top n markers based on log-fold changing values,we generated an expression heatmap for given cells and genesin the clusters.

Single cell trajectory analysisTrajectory analysis of scRNA-seq was performed using Monocle v.2.8.0(Qiu et al., 2017). First, datasets containing information of celldifferentiation were created using cell IDs selected through Seuratpackage. Then, we performed dimensional reduction using the DDRTreemethod to visualize the dataset, ordered the cells by global gene expressionlevels, and visualized the trajectory of the dataset. To illustrate the trends ofgene expression across pseudotime, we generated heatmaps to representsmooth expression curves of the genes.

Patterns of gene interactions during cell differentiationTo study gene interactions during cell differentiation, we used distancematrices of gene expressions. After selecting genes related to metabolicpathways, we collected the values of their gene expressions in each state andcalculated the similarities of the gene expressions using Pearson’s correlationcoefficient to generate the corresponding heatmaps (Stuart et al., 2003).

Single cell western blotFreshly FACS-isolatedMuSCs (1×105) were resuspended in 1 ml of loadingbuffer (ProteinSimple) and loaded on small scWest chip run (Milo System,ProteinSimple). MuSCs were settled for 15 min, followed by 5 s lysis and1 min separation (Milo System, ProteinSimple). scWest chip was incubatedwith sheep anti-histone H3 (Novus Biologicals, NB100-747; 1:40) anddonkey anti-rabbit IgG Alexa 555 (Invitrogen, A-31527; 1:40). Mouse anti-MyoD (Novus Biologicals, NBP2-32883) (100 µg/ml) was revealed bydonkey anti-mouse IgGAlexa 647 (Invitrogen, A-31571; 1:20). scWest chipwas scanned using a microarray scanner and analysis was performed withScout Software (Milo System, ProteinSimple).

AcknowledgementsWe thank Berendia Jackson, Dr Khooshbu Shah and the ProteinSimple team forassistance with single cell western blot analysis and Jim Simone from the FlowCytometry Section, NIAMS, for assistance with FACS. This study utilized the high-performance computational capabilities of the Helix System at the NIH, Bethesda,MD, USA (https://hpc.nih.gov/systems/helix.html/).

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: S.D., A.H.J., V.S.; Methodology: S.D., A.H.J., K.-D.K., F.N.,G.G.-C., X.F., V.S.; Software: K.-D.K.; Validation: A.H.J., K.-D.K., F.N., J.P., V.S.;Formal analysis: S.D., A.H.J., K.-D.K., F.N., V.S.; Investigation: S.D., A.H.J., F.N.,G.G.-C., X.F.; Resources: V.S.; Data curation: S.D., A.H.J., K.-D.K., X.F., V.S.;Writing - original draft: S.D., V.S.; Visualization: S.D., K.-D.K.; Supervision: V.S.;Project administration: V.S.; Funding acquisition: V.S.

FundingThis work was supported by the National Institute of Arthritis and Musculoskeletaland Skin Diseases Intramural Research Program (AR041126 and AR041164).Deposited in PMC for release after 12 months.

Data availabilityDatasets generated in this study are available in the Gene Expression Omnibusrepository under accession number GSE126834.

Supplementary informationSupplementary information available online athttp://dev.biologists.org/lookup/doi/10.1242/dev.174177.supplemental

ReferencesAguilar, C. A., Pop, R., Shcherbina, A., Watts, A., Matheny, R. W., Jr,

Cacchiarelli, D., Han, W. M., Shin, E., Nakhai, S. A., Jang, Y. C. et al. (2016).Transcriptional and chromatin dynamics of muscle regeneration after severetrauma. Stem Cell Rep. 7, 983-997.

Alonso-Martin, S., Rochat, A., Mademtzoglou, D., Morais, J., de Reynies, A.,Aurade, F., Chang, T. H.-T., Zammit, P. S. and Relaix, F. (2016). Geneexpression profiling of muscle stem cells identifies novel regulators of postnatalmyogenesis. Front. Cell Dev. Biol. 4, 58.

Andrews, J. L., Zhang, X., McCarthy, J. J., McDearmon, E. L., Hornberger, T. A.,Russell, B., Campbell, K. S., Arbogast, S., Reid, M. B., Walker, J. R. et al.(2010). CLOCK and BMAL1 regulate MyoD and are necessary for maintenance ofskeletal muscle phenotype and function. Proc. Natl. Acad. Sci. USA 107,19090-19095.

Beauchamp, J. R., Heslop, L., Yu, D. S. W., Tajbakhsh, S., Kelly, R. G., Wernig,A., Buckingham,M. E., Partridge, T. A. and Zammit, P. S. (2000). Expression ofCD34 and Myf5 defines the majority of quiescent adult skeletal muscle satellitecells. J. Cell Biol. 151, 1221-1234.

Bernet, J. D., Doles, J. D., Hall, J. K., Kelly Tanaka, K., Carter, T. A. and Olwin,B. B. (2014). p38 MAPK signaling underlies a cell-autonomous loss of stem cellself-renewal in skeletal muscle of aged mice. Nat. Med. 20, 265-271.

Boonsanay, V., Zhang, T., Georgieva, A., Kostin, S., Qi, H., Yuan, X., Zhou, Y.and Braun, T. (2016). Regulation of skeletal muscle stem cell quiescence bySuv4-20h1-Dependent facultative heterochromatin formation. Cell Stem Cell 18,229-242.

Brack, A. S. and Rando, T. A. (2007). Intrinsic changes and extrinsic influences ofmyogenic stem cell function during aging. Stem Cell Rev. 3, 226-237.

Brack, A. S. and Rando, T. A. (2012). Tissue-specific stem cells: lessons from theskeletal muscle satellite cell. Cell Stem Cell 10, 504-514.

Brack, A. S., Conboy, M. J., Roy, S., Lee, M., Kuo, C. J., Keller, C. and Rando,T. A. (2007). IncreasedWnt signaling during aging alters muscle stem cell fate andincreases fibrosis. Science 317, 807-810.

Buckingham,M. andRelaix, F. (2007). The role of Pax genes in the development oftissues and organs: Pax3 and Pax7 regulate muscle progenitor cell functions.Annu. Rev. Cell Dev. Biol. 23, 645-673.

Butler, A., Hoffman, P., Smibert, P., Papalexi, E. and Satija, R. (2018). Integratingsingle-cell transcriptomic data across different conditions, technologies, andspecies. Nat. Biotechnol. 36, 411-420.

Cacchiarelli, D., Qiu, X., Srivatsan, S., Manfredi, A., Ziller, M., Overbey, E.,Grimaldi, A., Grimsby, J., Pokharel, P., Livak, K. J. et al. (2018). Aligningsingle-cell developmental and reprogramming trajectories identifies moleculardeterminants of myogenic reprogramming outcome. Cell Syst. 7, 258-268.e253.

Chakkalakal, J. V., Jones, K. M., Basson,M. A. andBrack, A. S. (2012). The agedniche disrupts muscle stem cell quiescence. Nature 490, 355-360.

Chang, N. C., Sincennes, M. C., Chevalier, F. P., Brun, C. E., Lacaria, M.,Segales, J., Munoz-Canoves, P., Ming, H. and Rudnicki, M. A. (2018). Thedystrophin glycoprotein complex regulates the epigenetic activation of musclestem cell commitment. Cell Stem Cell 22, 755-768.e756.

Charge, S. B. and Rudnicki, M. A. (2004). Cellular and molecular regulation ofmuscle regeneration. Physiol. Rev. 84, 209-238.

Cheung, T. H., Quach, N. L., Charville, G.W., Liu, L., Park, L., Edalati, A., Yoo, B.,Hoang, P. andRando, T. A. (2012). Maintenance ofmuscle stem-cell quiescenceby microRNA-489. Nature 482, 524-528.

Cho, D. S. and Doles, J. D. (2017). Single cell transcriptome analysis of musclesatellite cells reveals widespread transcriptional heterogeneity. Gene 636, 54-63.

Collins, C. A., Olsen, I., Zammit, P. S., Heslop, L., Petrie, A., Partridge, T. A. andMorgan, J. E. (2005). Stem cell function, self-renewal, and behavioralheterogeneity of cells from the adult muscle satellite cell niche. Cell 122, 289-301.

Colombo, E., Bonetti, P., Lazzerini Denchi, E., Martinelli, P., Zamponi, R.,Marine, J.-C., Helin, K., Falini, B. and Pelicci, P. G. (2005). Nucleophosmin isrequired for DNA integrity and p19Arf protein stability. Mol. Cell. Biol. 25,8874-8886.

Comai, G. and Tajbakhsh, S. (2014). Molecular and cellular regulation of skeletalmyogenesis. Curr. Top. Dev. Biol. 110, 1-73.

Cosgrove, B. D., Gilbert, P. M., Porpiglia, E., Mourkioti, F., Lee, S. P., Corbel,S. Y., Llewellyn, M. E., Delp, S. L. and Blau, H. M. (2014). Rejuvenation of themuscle stem cell population restores strength to injured aged muscles. Nat. Med.20, 255-264.

Crist, C. G., Montarras, D. and Buckingham, M. (2012). Muscle satellite cells areprimed for myogenesis but maintain quiescencewith sequestration ofMyf5mRNAtargeted by microRNA-31 in mRNP granules. Cell Stem Cell 11, 118-126.

10

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

Das, S., Morvan, F., Morozzi, G., Jourde, B., Minetti, G. C., Kahle, P., Rivet, H.,Brebbia, P., Toussaint, G., Glass, D. J. et al. (2017). ATP citrate lyase regulatesmyofiber differentiation and increases regeneration by altering histone acetylation.Cell Rep. 21, 3003-3011.

de Morree, A., van Velthoven, C. T. J., Gan, Q., Salvi, J. S., Klein, J. D. D.,Akimenko, I., Quarta, M., Biressi, S. and Rando, T. A. (2017). Staufen1 inhibitsMyoD translation to actively maintain muscle stem cell quiescence. Proc. Natl.Acad. Sci. USA 114, E8996-E9005.

Dumont, N. A., Wang, Y. X., von Maltzahn, J., Pasut, A., Bentzinger, C. F., Brun,C. E. and Rudnicki, M. A. (2015). Dystrophin expression in muscle stem cellsregulates their polarity and asymmetric division. Nat. Med. 21, 1455-1463.

Faralli, H., Wang, C., Nakka, K., Benyoucef, A., Sebastian, S., Zhuang, L., Chu,A., Palii, C. G., Liu, C., Camellato, B. et al. (2016). UTX demethylase activity isrequired for satellite cell-mediated muscle regeneration. J. Clin. Invest. 126,1555-1565.

Feige, P., Brun, C. E., Ritso, M. andRudnicki, M. A. (2018). Orientingmuscle stemcells for regeneration in homeostasis, aging, and disease. Cell Stem Cell 23,653-664.

Fry, C. S., Kirby, T. J., Kosmac, K., McCarthy, J. J. and Peterson, C. A. (2017).Myogenic progenitor cells control extracellular matrix production by fibroblastsduring skeletal muscle hypertrophy. Cell Stem Cell 20, 56-69.

Garcıa-Prat, L., Martınez-Vicente, M., Perdiguero, E., Ortet, L., Rodrıguez-Ubreva, J., Rebollo, E., Ruiz-Bonilla, V., Gutarra, S., Ballestar, E., Serrano,A. L. et al. (2016). Autophagy maintains stemness by preventing senescence.Nature 529, 37-42.

Garcıa-Prat, L., Sousa-Victor, P. and Mun oz-Canoves, P. (2017). Proteostaticand metabolic control of stemness. Cell Stem Cell 20, 593-608.

Giordani, L., He, G. J., Negroni, E., Sakai, H., Law, J. Y. C., Siu, M. M., Wan, R.,Tajbakhsh, S., Cheung, T. H. and Le Grand, F (2018). High-dimensional single-cell cartography reveals novel skeletal muscle resident cell populations. bioRxivdoi:10.1101/304683

Hartl, F. U., Bracher, A. and Hayer-Hartl, M. (2011). Molecular chaperones inprotein folding and proteostasis. Nature 475, 324-332.

Hausburg, M. A., Doles, J. D., Clement, S. L., Cadwallader, A. B., Hall, M. N.,Blackshear, P. J., Lykke-Andersen, J. and Olwin, B. B. (2015). Post-transcriptional regulation of satellite cell quiescence by TTP-mediated mRNAdecay. eLife 4, e03390.

Janssen, I., Heymsfield, S. B., Wang, Z. M. and Ross, R. (2000). Skeletal musclemass and distribution in 468 men and women aged 18-88 yr. J. Appl. Physiol. 89,81-88.

Juan, A. H., Derfoul, A., Feng, X., Ryall, J. G., Dell’Orso, S., Pasut, A., Zare, H.,Simone, J. M., Rudnicki, M. A. and Sartorelli, V. (2011). Polycomb EZH2controls self-renewal and safeguards the transcriptional identity of skeletal musclestem cells. Genes Dev. 25, 789-794.

Kawabe, Y., Wang, Y. X., McKinnell, I. W., Bedford, M. T. and Rudnicki, M. A.(2012). Carm1 regulates Pax7 transcriptional activity through MLL1/2 recruitmentduring asymmetric satellite stem cell divisions. Cell Stem Cell 11, 333-345.

Kuang, S., Kuroda, K., Le Grand, F. and Rudnicki, M. A. (2007). Asymmetric self-renewal and commitment of satellite stem cells in muscle. Cell 129, 999-1010.

Liu, L., Cheung, T. H., Charville, G. W., Hurgo, B. M. C., Leavitt, T., Shih, J.,Brunet, A. and Rando, T. A. (2013). Chromatin modifications as determinants ofmuscle stem cell quiescence and chronological aging. Cell Rep. 4, 189-204.

Liu, L., Cheung, T. H., Charville, G. W. and Rando, T. A. (2015). Isolation ofskeletal muscle stem cells by fluorescence-activated cell sorting. Nat. Protoc. 10,1612-1624.

Lowe, M., Lage, J., Paatela, E., Munson, D., Hostager, R., Yuan, C., Katoku-Kikyo, N., Ruiz-Estevez, M., Asakura, Y., Staats, J. et al. (2018). Cry2 is criticalfor circadian regulation of myogenic differentiation by Bclaf1-mediated mRNAstabilization of Cyclin D1 and Tmem176b. Cell Rep. 22, 2118-2132.

Lukjanenko, L., Jung, M. J., Hegde, N., Perruisseau-Carrier, C., Migliavacca, E.,Rozo, M., Karaz, S., Jacot, G., Schmidt, M., Li, L. et al. (2016). Loss offibronectin from the aged stem cell niche affects the regenerative capacity ofskeletal muscle in mice. Nat. Med. 22, 897-905.

Machado, L., Esteves de Lima, J., Fabre, O., Proux, C., Legendre, R., Szegedi,A., Varet, H., Ingerslev, L. R., Barres, R., Relaix, F. et al. (2017). In situ fixationredefines quiescence and early activation of skeletal muscle stem cells. Cell Rep.21, 1982-1993.

McKinnell, I. W., Ishibashi, J., Le Grand, F., Punch, V. G., Addicks, G. C.,Greenblatt, J. F., Dilworth, F. J. and Rudnicki, M. A. (2008). Pax7 activatesmyogenic genes by recruitment of a histone methyltransferase complex.Nat. CellBiol. 10, 77-84.

Montarras, D., Morgan, J., Collins, C., Relaix, F., Zaffran, S., Cumano, A.,Partridge, T. and Buckingham, M. (2005). Direct isolation of satellite cells forskeletal muscle regeneration. Science 309, 2064-2067.

Nakahata, Y., Sahar, S., Astarita, G., Kaluzova,M. andSassone-Corsi, P. (2009).Circadian control of the NAD+ salvage pathway by CLOCK-SIRT1. Science 324,654-657.

Nguyen, V.-A. and Lio, P. (2009). Measuring similarity between gene expressionprofiles: a Bayesian approach. BMC Genomics 10 Suppl. 3, S14.

Oustanina, S., Hause, G. and Braun, T. (2004). Pax7 directs postnatal renewal andpropagation of myogenic satellite cells but not their specification. EMBO J. 23,3430-3439.

Pala, F., Di Girolamo, D., Mella, S., Yennek, S., Chatre, L., Ricchetti, M. andTajbakhsh, S. (2018). Distinct metabolic states govern skeletal muscle stem cellfates during prenatal and postnatal myogenesis. J. Cell Sci. 131, jcs212977.

Pallafacchina, G., François, S., Regnault, B., Czarny, B., Dive, V., Cumano, A.,Montarras, D. and Buckingham, M. (2010). An adult tissue-specific stem cell inits niche: a gene profiling analysis of in vivo quiescent and activated musclesatellite cells. Stem Cell Res. 4, 77-91.

Price, F. D., von Maltzahn, J., Bentzinger, C. F., Dumont, N. A., Yin, H., Chang,N. C., Wilson, D. H., Frenette, J. and Rudnicki, M. A. (2014). Inhibition of JAK-STAT signaling stimulates adult satellite cell function. Nat. Med. 20, 1174-1181.

Puri, P. L. and Mercola, M. (2012). BAF60 A, B, and Cs of muscle determinationand renewal. Genes Dev. 26, 2673-2683.

Qiu, X., Mao, Q., Tang, Y., Wang, L., Chawla, R., Pliner, H. A. and Trapnell, C.(2017). Reversed graph embedding resolves complex single-cell trajectories.Nat.Methods 14, 979-982.

Rocheteau, P., Gayraud-Morel, B., Siegl-Cachedenier, I., Blasco, M. A. andTajbakhsh, S. (2012). A subpopulation of adult skeletal muscle stem cells retainsall template DNA strands after cell division. Cell 148, 112-125.

Rodgers, J. T., King, K. Y., Brett, J. O., Cromie, M. J., Charville, G. W., Maguire,K. K., Brunson, C., Mastey, N., Liu, L., Tsai, C. R. et al. (2014). mTORC1controls the adaptive transition of quiescent stem cells from G0 to G(Alert). Nature510, 393-396.

Ryall, J. G., Cliff, T., Dalton, S. and Sartorelli, V. (2015a). Metabolicreprogramming of stem cell epigenetics. Cell Stem Cell 17, 651-662.

Ryall, J. G., Dell’Orso, S., Derfoul, A., Juan, A., Zare, H., Feng, X., Clermont, D.,Koulnis, M., Gutierrez-Cruz, G., Fulco, M. et al. (2015b). The NAD(+)-dependent SIRT1 deacetylase translates a metabolic switch into regulatoryepigenetics in skeletal muscle stem cells. Cell Stem Cell 16, 171-183.

Sacco, A. andPuri, P. L. (2015). Regulation ofmuscle satellite cell function in tissuehomeostasis and aging. Cell Stem Cell 16, 585-587.

Sacco, A., Doyonnas, R., Kraft, P., Vitorovic, S. and Blau, H. M. (2008). Self-renewal and expansion of single transplanted muscle stem cells. Nature 456,502-506.

Satija, R., Farrell, J. A., Gennert, D., Schier, A. F. and Regev, A. (2015). Spatialreconstruction of single-cell gene expression data. Nat. Biotechnol. 33, 495-502.

Scionti, I., Hayashi, S., Mouradian, S., Girard, E., Esteves de Lima, J., Morel, V.,Simonet, T., Wurmser, M., Maire, P., Ancelin, K. et al. (2017). LSD1 controlstimely MyoD expression via MyoD core enhancer transcription. Cell Rep. 18,1996-2006.

Seale, P., Sabourin, L. A., Girgis-Gabardo, A., Mansouri, A., Gruss, P. andRudnicki, M. A. (2000). Pax7 is required for the specification of myogenic satellitecells. Cell 102, 777-786.

Sherwood, R. I., Christensen, J. L., Conboy, I. M., Conboy, M. J., Rando, T. A.,Weissman, I. L. and Wagers, A. J. (2004). Isolation of adult mouse myogenicprogenitors: functional heterogeneity of cells within and engrafting skeletalmuscle. Cell 119, 543-554.

Shi, S. Y., Lu, S. Y., Sivasubramaniyam, T., Revelo, X. S., Cai, E. P., Luk, C. T.,Schroer, S. A., Patel, P., Kim, R. H., Bombardier, E. et al. (2015). DJ-1 linksmuscle ROS production with metabolic reprogramming and systemic energyhomeostasis in mice. Nat. Commun. 6, 7415.

Sincennes, M.-C., Brun, C. E. and Rudnicki, M. A. (2016). Concise review:epigenetic regulation of myogenesis in health and disease. Stem Cells Transl.Med. 5, 282-290.

Solanas, G., Peixoto, F. O., Perdiguero, E., Jardi, M., Ruiz-Bonilla, V., Datta, D.,Symeonidi, A., Castellanos, A., Welz, P. S., Caballero, J. M. et al. (2017). Agedstem cells reprogram their daily rhythmic functions to adapt to stress. Cell 170,678-692.e620.

Sousa-Victor, P., Gutarra, S., Garcıa-Prat, L., Rodriguez-Ubreva, J., Ortet, L.,Ruiz-Bonilla, V., Jardi, M., Ballestar, E., Gonzalez, S., Serrano, A. L. et al.(2014). Geriatric muscle stem cells switch reversible quiescence into senescence.Nature 506, 316-321.

Spalding, K. L., Bhardwaj, R. D., Buchholz, B. A., Druid, H. and Frisen, J. (2005).Retrospective birth dating of cells in humans. Cell 122, 133-143.

Stuart, J. M., Segal, E., Koller, D. and Kim, S. K. (2003). A gene-coexpressionnetwork for global discovery of conserved genetic modules. Science 302,249-255.

Tabula Muris Consortium et al. (2018). Single-cell transcriptomics of 20 mouseorgans creates a Tabula Muris. Nature 562, 367-372.

Tang, A. H. and Rando, T. A. (2014). Induction of autophagy supports thebioenergetic demands of quiescent muscle stem cell activation. EMBO J. 33,2782-2797.

Tierney, M. T., Aydogdu, T., Sala, D., Malecova, B., Gatto, S., Puri, P. L., Latella,L. and Sacco, A. (2014). STAT3 signaling controls satellite cell expansion andskeletal muscle repair. Nat. Med. 20, 1182-1186.

Tosic, M., Allen, A., Willmann, D., Lepper, C., Kim, J., Duteil, D. and Schule, R.(2018). Lsd1 regulates skeletal muscle regeneration and directs the fate ofsatellite cells. Nat. Commun. 9, 366.

11

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT

Trapnell, C., Cacchiarelli, D., Grimsby, J., Pokharel, P., Li, S., Morse, M.,Lennon, N. J., Livak, K. J., Mikkelsen, T. S. and Rinn, J. L. (2014). Thedynamics and regulators of cell fate decisions are revealed by pseudotemporalordering of single cells. Nat. Biotechnol. 32, 381-386.

van den Brink, S. C., Sage, F., Vertesy, A., Spanjaard, B., Peterson-Maduro, J.,Baron, C. S., Robin, C. and van Oudenaarden, A. (2017). Single-cellsequencing reveals dissociation-induced gene expression in tissuesubpopulations. Nat. Methods 14, 935-936.

van Velthoven, C. T. J., de Morree, A., Egner, I. M., Brett, J. O. and Rando, T. A.(2017). Transcriptional profiling of quiescent muscle stem cells in vivo. Cell Rep.21, 1994-2004.

Verma, M., Asakura, Y., Murakonda, B. S. R., Pengo, T., Latroche, C., Chazaud,B., McLoon, L. K. and Asakura, A. (2018). Muscle satellite cell cross-talk with avascular niche maintains quiescence via VEGF and notch signaling. Cell StemCell 23, 530-543.e539.

White, R. B., Bierinx, A.-S., Gnocchi, V. F. and Zammit, P. S. (2010). Dynamics ofmuscle fibre growth during postnatal mouse development. BMC Dev. Biol. 10, 21.

Yennek, S., Burute, M., Thery, M. and Tajbakhsh, S. (2014). Cell adhesiongeometry regulates non-random DNA segregation and asymmetric cell fates inmouse skeletal muscle stem cells. Cell Rep. 7, 961-970.

Zammit, P. S., Heslop, L., Hudon, V., Rosenblatt, J. D., Tajbakhsh, S.,Buckingham, M. E., Beauchamp, J. R. and Partridge, T. A. (2002). Kineticsof myoblast proliferation show that resident satellite cells are competent to fullyregenerate skeletal muscle fibers. Exp. Cell Res. 281, 39-49.

Zhang, X., Li, T., Liu, F., Chen, Y., Yao, J., Li, Z., Huang, Y. and Wang, J. (2019).Comparative analysis of droplet-based ultra-high-throughput single-cell RNA-seqsystems. Mol. Cell 73, 130-142.e5.

Zheng, J., Lu, Z., Kocabas, F., Bottcher, R. T., Costell, M., Kang, X., Liu, X.,Deberardinis, R. J., Wang, Q., Chen, G.-Q. et al. (2014). Profilin 1 is essential forretention and metabolism of mouse hematopoietic stem cells in bone marrow.Blood 123, 992-1001.

Zismanov, V., Chichkov, V., Colangelo, V., Jamet, S., Wang, S., Syme, A.,Koromilas, A. E. and Crist, C. (2016). Phosphorylation of eIF2alpha is atranslational control mechanism regulating muscle stem cell quiescence and self-renewal. Cell Stem Cell 18, 79-90.

12

TECHNIQUES AND RESOURCES Development (2019) 146, dev174177. doi:10.1242/dev.174177

DEVELO

PM

ENT